A Comparative Study of Different Automated Segmentation Techniques for Brain Tumor

Brain tumor detection is one of the challenging tasks in medical field and digital image processing. A tumor is the abnormal growth of cells in human brain. Such abnormal cell growth may cause a cancer in the human brain which according to a recent survey has been one of the most rapidly increasing reasons of deaths. With such an alarming number of cancer patients due to brain tumor, it is important to automate the diagnosis process for accurate and fast detection of tumors in the MRI scan of the patients. This research work thus discusses in detail the different automated segmentation techniques for detection and segmentation of tumor in MR images of brain. A brain tumor dataset of Surgical Planning Laboratory (SPL) and Department of Neurosurgery (NSG) is used which contains MRI brain tumor images of ten patients. The purpose of this research work is to compare different segmentation technique like threshold segmentation, histogram, watershed segmentation and k-means clustering and to find strategy which gives most accurate results. Segmentation techniques followed by morphological operations with optimum disk size of structuring element is used. The accuracy of Watershed technique using our segmentation strategy is 99.86% which is best of all other selected segmentation algorithms.